Genetic parameters for novel behavioral traits for cows housed under automatic milking systems with a focus on learning.

Abstract

Automatic milking systems (AMS) are increasingly used in dairy farming, highlighting the importance that cows adapt to this technical environment. This study characterized novel AMS traits related to learning behavior and milking temperament and evaluated their phenotypic and genetic properties to assess their suitability for breeding. It was investigated whether pre-lactational familiarization to AMS is related to cow-robot-performance and behavior during 1st parity and to what extent cow-AMS-interactions develop across parities (1st - 5^th+ parity). AMS-derived traits rejected milkings (RM) and visiting intervals (VI) were used to investigate learning behavior, i.e., cows' ability to voluntarily visit the robot, while udder scan time (UST) and connecting time (CT) were used to characterize temperament behavior inside the AMS. Data comprised of 17,288 Holstein cows with 41,581 lactations from 24 German dairy farms. To assess pre-lactational familiarization effects, animals were grouped by the number of pre-lactational visits to the AMS: group 0 (none, n = 14,840), group 1 (1-7 visits before 1st parity at AMS, n = 1,294), group 2 (>7 visits before 1st parity at AMS, n = 1,154). It could be shown that pre-lactational familiarization influenced later AMS performance: intensive familiarization (group 2) resulted in high RM and short VI, whereas cows with minimal familiarization (group 1) showed the lowest RM. Cows without familiarization (group 0) exhibited decreasing RM with increasing milking experience. UST showed negligible differences between groups, while CT was shortest in non-familiarized cows. To investigate the cow-AMS-interactions across parities, only non-familiarized cows (group 0) were considered. Across parities, clear learning and age effects were observed. Primiparous cows exhibited the highest RM (LSM = 3.09 ± 0.04) and longest VI (LSM = 389.59 ± 1.78 min), with limited improvement during the first 99 d in milk. Cows in 2nd parity adapted more rapidly, showing a decrease in RM from 2.94 ± 0.05 to 1.58 ± 0.06 during the first 99 d as well as a decrease in VI from 366.23 ± 2.22 min to 352.56 ± 2.57 min. RM further declined consistently across the 2nd to 4th parity, indicating both short-term and long-term learning effects. UST and CT were highest in early lactation and declined during the first 100 d, reflecting increasing familiarity with the milking process. Differences between parities were small but consistent: older cows required less time for UST and CT. Genetic analyses revealed heritabilities of 0.1 ± 0.005 for RM and 0.25 ± 0.007 for VI, with a highly negative genetic (-0.98) and phenotypic (-0.81) correlation between them. UST and CT exhibited low heritabilities (0.04 ± 0.003 and 0.07 ± 0.005), indicating stronger environmental influence. A positive genetic correlation (0.76 ± 0.06) suggests underlying temperament that may influence AMS efficiency. Overall, AMS-derived behavioral and temperament traits capture complementary aspects of cow-robot interaction. Particularly VI and RM exhibit exploitable genetic variation and may serve as informative indicators for improving voluntary attendance, cow traffic, and overall AMS efficiency. Integrating such traits into breeding programs may support precision dairy management and enhance the AMS performance.